Backtesting ArenaBacktesting Arena
← Back to blog

Robbed! (2) The Biggest Crypto Hacks – and When Your Money Is Really Gone

The biggest crypto hack in history cost Bybit about 1.5 billion dollars in February 2025. Its customers lost nothing. Thieves give back almost nothing: 3.1 percent of everything stolen. So whether you see your money again depends on who can carry the loss. A solvent exchange can. A bankrupt one cannot, and in your own wallet, it is you. The biggest hacks, how they worked, and what follows for where you keep your coins.

Backtesting Arena·October 19, 2026·12 min read·3 views
Robbed! (2) The Biggest Crypto Hacks – and When Your Money Is Really Gone

On 21 February 2025, about 1.5 billion dollars disappeared from the exchange Bybit, including 401,347 ETH. It is the largest crypto theft ever. Five days later, the FBI named the culprit: North Korea.

And the customers? They lost nothing. Bybit covered the loss in less than 72 hours with loans and its own money.

If you are wondering where to keep your coins, this is the most important lesson from the big hacks. Hacks happen everywhere: on exchanges, in bridges, in DeFi and in private wallets. Whether you see your money again depends on who can carry the loss.

Part 1 of this series was about fraud in which the operators themselves were the culprits. This part is about theft by outsiders.

The biggest hacks: the sums grow, customers often got away

This is the list of the biggest hacks, sorted by dollar value on the day of the theft:

CaseDatethenthe coins todayWhereWhat customers got back
Bybit21 Feb 2025~$1.5bn$1.0bn (ETH only)exchangeeverything, from Bybit
Ronin Bridge23 Mar 2022~$624m$458mbridgenot documented
Poly Network10 Aug 2021$611m–bridgeeverything, from the attacker
BNB Bridge6 Oct 2022~$570m$1.48bnbridge$100–137m escaped, the rest stayed frozen
Coincheck26 Jan 2018~$530m$0.27mexchangecompensation in yen, from Coincheck
Mt. Gox (borderline)to Feb 2014$470m$70.7bnexchangesince July 2024, in BTC, after ten years
Bitget24 Sep 2026up to $388m–exchangeprotection fund announced
Wormhole2 Feb 2022~$320m$299mbridgeeverything, from investor Jump Crypto
Liquid Network6 Sep 2026~$320m$333msidechain85 %, from the attacker
DMM Bitcoin31 May 2024~$305m$375mexchangeeverything, from DMM; the exchange closed
Drift1 Apr 2026~$285–295m–DeFinothing from the attacker
Kelp rsETH18 Apr 2026~$291m~$290mbridgejust over a quarter frozen
KuCoin25 Sep 2020~$281–285m–exchange84 % recovered, the rest from an insurance fund
WazirX18 Jul 2024~$235m–exchangeinsolvency proceedings, payout rate disputed
Cetus22 May 2025$223m–DeFi73 % frozen, the rest from Cetus and the Sui Foundation

DefiLlama, FBI, company statements; "today" = the stolen coin amount at the price of 13 Oct 2026, only where the amount is documented. Mt. Gox is a borderline case because the theft mixed with poor bookkeeping over years. The figure of 850,000 BTC is the standard number from the time; 200,000 BTC turned up again later.

Three things stand out. Six of the 15 cases are from 2025 and 2026. Counted in coins, Mt. Gox is by far the largest case; the coins would be worth 70.7 billion dollars today. And in most of the big cases, customers got all or almost all of their money back.

The biggest hack is not the biggest loss for customers. That is decided afterwards.

So where does the money come from, if not from the thief?

The thieves give back almost nothing

DefiLlama counts 1,301 hacks since 2011. In 27 of them, money came back from the attacker. That is 3.1 percent of everything stolen.

For North Korea, that figure is practically zero. Of the Bybit loot, about 43 million dollars were frozen, roughly 3 percent. And since 2022, North Korea has been the largest single culprit. According to Chainalysis, it accounted for:

  • 2021: 12 percent of the stolen sums
  • 2022: 45 percent
  • 2024: 61 percent
  • 2025: 59 percent, about 2 of 3.4 billion dollars

For 2026, private analytics firms also attribute Bitget, Drift and Kelp to North Korea. There is no official FBI confirmation of that to date.

When customers did see money, it came from three sources:

  1. The operator paid out of its own pocket. Bybit, Coincheck and DMM covered the loss with their own capital, KuCoin also with an insurance fund.
  2. An investor or the chain stepped in. At Wormhole, the investor Jump Crypto replaced 120,000 ETH the day after. At Cetus, the validators of the Sui chain froze 162 million dollars. At the BNB Bridge, the validators halted the entire chain for about three hours.
  3. The attacker gave it back. At Poly Network everything came back, at Euler about 90 percent, at Liquid Network 85 percent. This happened almost only with attackers who were not state hackers and were being tracked in public.

You cannot count on the thief. You can only count on whoever stands between you and the thief.

How do the attackers get at so much money in the first place?

The way in is usually a person with a key

Anyone who wants to move coins needs a private key. For large holdings it often takes several: a multisig wallet only releases a payment once, say, three of five key holders sign. These people are called signers.

That is exactly where the biggest hacks strike. According to DefiLlama, 49 percent of all stolen sums came from stolen keys or deceived people. In 2024 it was 69 percent, in 2025 68 percent.

This is how it went in the big cases:

  • Bybit: attackers compromised the computer of a developer at Safe, the provider of the multisig interface. The Bybit signers saw a harmless payment on screen. In fact, they handed over access to their ETH reserve.
  • DMM Bitcoin: a fake recruiter sent an employee of the wallet provider Ginco a "hiring test". Through it, the attackers got hold of that employee's session and manipulated a real transaction. That is how the FBI describes it.
  • Ronin Bridge: five of nine keys were enough to release funds. Four of them were held by the same company, Sky Mavis. The attackers got the fifth through an old approval that had never been revoked.
  • Drift: the attackers got two of five members of a security council to sign transactions in advance. After a restructuring, there was no longer a waiting period in which such a thing could have been stopped.

A bridge, by the way, connects two blockchains. You hand over coins on one side and get a counterpart on the other. For that, a large pot sits on the first side, and whoever holds the keys to that pot holds everything.

It is rarely the code that gets hacked. It is the person who signs.

And where is the money that disappears this way?

The risk moves, it does not go away

Which type of target lost the most money has flipped several times over the years. Shares of each year's total according to DefiLlama:

YearExchangesBridgesDeFi
201895 %–3 %
202114 %23 %63 %
202216 %52 %26 %
202570 %1 %27 %
2026 so far18 %34 %35 %

DefiLlama hacks, own calculation. DefiLlama barely records thefts from private wallets; Chainalysis is used for those below.

Until 2019, almost only exchanges lost. In 2021 it was DeFi, in 2022 the bridges, in 2025 the exchanges again. In 2026 it is DeFi and bridges once more. Attackers go where a lot of money currently sits behind few keys.

On top of that, the number of hacks is rising, but the money hangs on a few big ones. For 2026, DefiLlama already counts 294 cases. The median case was only 0.45 million dollars, against 3.4 million in 2021. The 42 hacks of 100 million dollars or more account for 64 percent of all stolen sums.

There is no place that never gets hacked. There are only places where someone can pay for the damage.

So who can?

Who pays when it happens

Where your coins sitWho carries the lossWhat the big cases show
an exchange that can absorb the hackthe exchangeBybit, Coincheck, DMM, KuCoin: customers compensated, often within days
an exchange that cannot absorb the hackyou, through insolvency proceedingsMt. Gox: ten years until repayment. WazirX: about 45 % of reserves gone, payout rate disputed
a bridgesometimes an investor, otherwise the loss stays with usersWormhole replaced, Poly returned; at the BNB Bridge $100–137m escaped
a DeFi protocolthe depositors, sometimes the protocolEuler 90 % back, Cetus through freezing and loans; Kelp also hit the lending market Aave with a $123–230m shortfall
your own walletyouChainalysis 2025: 158,000 incidents, at least 80,000 victims

So with an exchange, what counts is not whether it can be hacked. Any can. What counts is whether it can still pay afterwards. Bybit lost about 1.5 billion dollars and stayed solvent. WazirX lost about 235 million, roughly 45 percent of its reserves, and ended up in insolvency proceedings.

The size of the hack matters less than the size of the cushion behind it.

Is your own wallet the better choice, then?

"Not your keys, not your coins": your own wallet moves the risk to you

"Then I will just keep my coins in my own wallet. Then no exchange can lose them."

True, and your own wallet does protect you against an exchange going bust. But it shifts the risk to you. According to Chainalysis, private wallets accounted for about 20 percent of the value stolen in 2025, around 713 million dollars. It counted 158,000 incidents with at least 80,000 victims. In absolute numbers, that is where most victims are. And for them, no exchange, investor or insurance fund steps in.

The wallet itself is not always safe either. In July 2026, about 500 owners of the hardware wallet COLDCARD lost a combined 594 BTC, about 38 million dollars. The cause was a flaw in generating the random numbers for the keys. That is the rare case in which the device itself was the weak point.

And the trick used to rob Bybit works on you too. The signers saw a harmless payment and signed something else. Anyone who uses a wallet signs all the time.

In your own wallet, you are the exchange, the signer and the insurance fund at once.

What does this mean for where you keep your coins?

Where you keep your coins: questions for the exchange and for yourself

If you use an exchange, do not ask about the name. Ask about the cushion:

  • Does it publish proof of reserves, that is, evidence that it really holds customers' coins? Bybit published one after the hack showing more than 100 percent.
  • Is there a protection or insurance fund, and who checks it? Bitget states more than 464 million dollars for its fund. That has not been independently verified.
  • How much is in it compared with what a big hack would cost? WazirX lost 45 percent of its reserves in one go.
  • What you cannot check: how the exchange manages its keys. That is the part the big hacks go after.

If you use a bridge or a DeFi protocol, ask about the keys:

  • How many signatures does a release need, and who holds them? At Kelp, a single verifier was enough. That was publicly visible in the configuration.
  • Is there a waiting period before changes? At Drift it was last set to zero seconds, also publicly visible.

If you use your own wallet, you are your own security department:

  • Read what you sign on the device, not just on the website. At Bybit, the website showed something other than what was signed.
  • Keep your recovery words separate from the device and never enter them anywhere someone sends you.
  • Spread large holdings instead of putting everything on one key.

These numbers cannot tell you which exchange or wallet is safe. They only show who carried the loss in the big cases. Whether the big cases of 2026 move back to exchanges or stay with bridges and DeFi, only the year-end reports will show.

FAQ

Why is Mt. Gox only a borderline case in the list? At Mt. Gox, coins disappeared over years, and the bookkeeping concealed it. Former CEO Mark Karpelès was convicted in Tokyo in 2019 only for manipulating data, not for the loss. Who took the coins has not been established in court. Creditors have been getting money back since July 2024. On 16 October 2026, the trustee extended the deadline for that again, from 31 October 2026 to 31 October 2027.

Why is LuBian missing, the largest hack according to DefiLlama? DefiLlama lists LuBian with 127,426 BTC from December 2020. According to the analytics firms Elliptic and Arkham, these are the same coins the US Department of Justice moved to seize in October 2025 as proceeds of the Prince Group fraud. Whether it was a theft by outsiders is open.

Why do the annual totals differ by source? Each provider counts differently. For 2025, Chainalysis gives more than 3.4 billion dollars, CertiK 3.35, SlowMist 2.94 and DefiLlama 2.72 billion. The range is about 25 percent. So always compare years using the same source.

What about Bitfinex 2016? At the time, 119,754 BTC were stolen, about 72 million dollars. Today they would be worth almost 10 billion dollars. Ilya Lichtenstein was sentenced in 2024 to five years for money laundering, Heather Morgan to 18 months. 94,643 BTC are to go back to Bitfinex. Whether that has happened is not documented to date.

How it was measured

Cases and sums from the DefiLlama hacks database (1,301 entries, as of 13 Oct 2026), supplemented by statements from the FBI, Japan's National Police Agency, company announcements and media. Annual figures and North Korea shares from Chainalysis (reports of 19 Dec 2024 and 18 Dec 2025); shares by target type and by way in as own calculations from DefiLlama, excluding LuBian. DefiLlama lists the full sum as "returned" for Bybit and DMM; what is meant there is customer compensation, not money from the attacker. That entry has been removed. "Today" values: stolen coin amounts at the price of 13 Oct 2026, 11:14 UTC. Culprits are named only where a court or an authority has established them.


Not investment advice, not a recommendation, not a forecast — historical patterns are no promise.

Study the Past — Improve your Future 🥋

Try it yourself

Run the backtest with your own parameters and time ranges.

Run backtest →

More on this topic

Market Analysis

Robbed! (1) The Biggest Crypto Scams – and How You Could Have Spotted Them Beforehand

Backtesting Arenatradingstrategies.work

OneCoin, Bitconnect, PlusToken, Celsius, FTX: billions gone, millions of victims. In 9 of the 15 biggest cases, the warning sign was in the offer itself, before anyone paid in: a fixed return, money for recruiting, a source of profit nobody could check. What the biggest scams had in common, how little victims got back, why the fraud of today looks different, and five questions to ask before you deposit.

Risk managementCryptoBitcoin
Oct 19, 20261 min
Market Analysis

Bitcoin's 2026 Fall Was One of Its Mildest – Yet It Sent the Price Further Back in Time Than Ever

Backtesting Arenatradingstrategies.work

Measured by drawdown, the 2026 bear market was the mildest of bitcoin's big cycles: −53 % against −77 % to −93 % before. Measured in time, it was the worst: the low matched a price from five and a quarter years earlier, further back than any low since 2010. And against its own history, bitcoin was not cheap at any low since 2017. Three numbers for every bear market, and what each tells you before you call bitcoin a bargain.

DrawdownBitcoinMethodology+1
Oct 15, 20261 min
Market Analysis

How Often Does Bitcoin Crash 30% – and Is It Getting Rarer?

Backtesting Arenatradingstrategies.work

Since 2011 Bitcoin has fallen 30 percent within a month 19 times. How often that happens today, how it fits with falling volatility, how far the drop usually went, and what that means for the size of your position.

DrawdownRisk managementBitcoin+1
Oct 5, 20261 min
Market Analysis

Altcoin Season in Sight: Is It Time to Switch Out of Bitcoin?

Backtesting Arenatradingstrategies.work

When half the top altcoins beat Bitcoin, an altcoin season usually follows. Yet in 11 of 13 cases since 2020, switching to altcoins meant falling behind Bitcoin, whether you bought the leaders or the rest. Five checks and what each one tells you.

MomentumMarket structureBitcoin+1
Oct 2, 20261 min
📬

Don't miss new blog posts

One short email per new post — strategies, backtests, market analysis. No spam, unsubscribe with one click anytime.

By subscribing you accept our privacy policy. We use Resend for delivery. Double opt-in confirmation required.

Comments (0)

Join free to post comments.

Sign up →

No comments yet. Be the first!